Mixed-phase clouds are important for simulating precipitation formation and cloud radiative effects in Numerical Weather Prediction (NWP) and climate models. One challenge to reduce model uncertainties is how to best represent the sub-grid inhomogeneity of liquid and ice within a model grid-box. This is poorly constrained by observations, yet is key for representing microphysical process rates that grow ice crystals at the expense of liquid droplets. This study uses in-situ airborne observations from stratiform, shallow cumulus, deep convective and frontal clouds to investigate the spatial overlap of liquid and ice phases on length scales ~1 km, which are appropriate for current regional NWP models. We place observational constraints on a simple parametrisation that describs the mixed-phase fraction as a function of sub-grid liquid and ice cloud fractions and demonstrate that most of the observations show that when ice and liquid are present they are close to fully overlapped.